Skip to contentSkip to conversation

kerneloops.org: 2.6.28-rc regression in epoll (list corruption)

DISCUSSION Arjan van de Ven 4 participants 6 messages Oct 26, 2008 VFS

lore
Share & tools
lore
#3 lore raw
In reply to Arjan van de Ven
Quoted 6 lines · Arjan van de Ven

This one is upcoming fast (and I just hit it as well)

http://www.kerneloops.org/searchweek.php?search=ep_poll_callback

seems epoll grew some list corruption....

It sounds very much like f337b9c58332bdecde965b436e47ea4c94d30da0 ("epoll:
drop unnecessary test") deleted a test that wasn't so unnecessary after all..

That ep_poll_callback() code is:

        /* If this file is already in the ready list we exit soon */
        if (ep_is_linked(&epi->rdllink))
                goto is_linked;

        list_add_tail(&epi->rdllink, &ep->rdllist);

and the unnecessary test that was removed looks _very_ much like that kind of code.

Thomas? Davide?

And if somebody knows how to reproduce this reliably, it would be really good to hear if doing a revert on that thing just fixed it. It should revert cleanly - it's the only change to fs/eventpoll.c since 2.6.27.

I'm somewhat inclined to revert it without even getting confirmation, since I wanted to do an early -rc2 today with all the brown-paper-bag fixes that have accumulated. But it would be good to get some confirmation.

Btw, that whole logic in ep_send_events() sounds a bit scary. It says:

We can loop without lock because this is a task private list.

and it's true that "txlist" is a private list, but it still seems to depend on the fact that none of the "struct epitem"s on that list can be reached by any other means. And that whole thing depends on the magic behaviour of 'ep->ovflist', but there are a lot of code sequences that do *not* seem to test that at all.

Maybe the rqace/bug has always been there (ie some sequence that works with an "epi->rdllink" without checking ovflist), but the "unnecessary" test protected us from seeing it in practice.

Linus

#4 lore raw
Quoted 25 lines · On Sun, 26 Oct 2008 11:58:06 -0700 (PDT) Linus Torvalds…
Quoted 6 lines · Arjan van de Ven

This one is upcoming fast (and I just hit it as well)

http://www.kerneloops.org/searchweek.php?search=ep_poll_callback

seems epoll grew some list corruption....

It sounds very much like f337b9c58332bdecde965b436e47ea4c94d30da0 ("epoll: drop unnecessary test") deleted a test that wasn't so unnecessary after all..

That ep_poll_callback() code is:

        /* If this file is already in the ready list we exit soon */
        if (ep_is_linked(&epi->rdllink))
                goto is_linked;

        list_add_tail(&epi->rdllink, &ep->rdllist);

and the unnecessary test that was removed looks _very_ much like that kind of code.

Quoted 7 lines

Thomas? Davide?

And if somebody knows how to reproduce this reliably, it would be really good to hear if doing a revert on that thing just fixed it. It should revert cleanly - it's the only change to fs/eventpoll.c since 2.6.27.

I did the revert and did the same thing I did before, and haven't seen it yet.

(this is obviously a really small sample size on both sides of the line, but it's at least not show it easily now while I saw it a few times before.. after I fixed my Wifi and manually set the r8169 mac)

Signature · 3 lines
Arjan van de Ven 	Intel Open Source Technology Centre
For development, discussion and tips for power savings, 
visit http://www.lesswatts.org
#5 lore raw
In reply to Linus Torvalds
Quoted 16 lines · Linus Torvalds
Quoted 6 lines · Arjan van de Ven

This one is upcoming fast (and I just hit it as well)

http://www.kerneloops.org/searchweek.php?search=ep_poll_callback

seems epoll grew some list corruption....

It sounds very much like f337b9c58332bdecde965b436e47ea4c94d30da0 ("epoll:
drop unnecessary test") deleted a test that wasn't so unnecessary after all..

That ep_poll_callback() code is:

Quoted 8 lines
        /* If this file is already in the ready list we exit soon */
        if (ep_is_linked(&epi->rdllink))
                goto is_linked;

        list_add_tail(&epi->rdllink, &ep->rdllist);

and the unnecessary test that was removed looks _very_ much like that kind of code.

No, the test was in the re-insertion loop. This is the patch that fixes it:

---
 fs/eventpoll.c |   11 +++++++++--
 1 file changed, 9 insertions(+), 2 deletions(-)
fs/eventpoll.c +9 -2
Index: linux-2.6.mod/fs/eventpoll.c
===================================================================
@@ -930,8 +930,15 @@
 	 * inside the main ready-list here.
 	 */
 	for (nepi = ep->ovflist; (epi = nepi) != NULL;
-	     nepi = epi->next, epi->next = EP_UNACTIVE_PTR)
-		list_add_tail(&epi->rdllink, &ep->rdllist);
+	     nepi = epi->next, epi->next = EP_UNACTIVE_PTR) {
+		/*
+		 * If the above loop quit with errors, the epoll item might still
+		 * be linked to "txlist", and the list_splice() done below will
+		 * take care of those cases.
+		 */
+		if (!ep_is_linked(&epi->rdllink))
+			list_add_tail(&epi->rdllink, &ep->rdllist);
+	}
 	/*
 	 * We need to set back ep->ovflist to EP_UNACTIVE_PTR, so that after
 	 * releasing the lock, events will be queued in the normal way inside
Quoted 3 lines

And if somebody knows how to reproduce this reliably, it would be really good to hear if doing a revert on that thing just fixed it. It should revert cleanly - it's the only change to fs/eventpoll.c since 2.6.27.

That's 100% due to the removed test that went in when you sucked up Andrew bits after .27.
Patch has been confirmed by me and bug submitters.

Quoted 4 lines

I'm somewhat inclined to revert it without even getting confirmation, since I wanted to do an early -rc2 today with all the brown-paper-bag fixes that have accumulated. But it would be good to get some confirmation.

No need to revert, since one of the removed tests is really un-needed.

Quoted 10 lines

Btw, that whole logic in ep_send_events() sounds a bit scary. It says:

We can loop without lock because this is a task private list.

and it's true that "txlist" is a private list, but it still seems to depend on the fact that none of the "struct epitem"s on that list can be reached by any other means. And that whole thing depends on the magic behaviour of 'ep->ovflist', but there are a lot of code sequences that do *not* seem to test that at all.

Quoted 3 lines

Maybe the rqace/bug has always been there (ie some sequence that works with an "epi->rdllink" without checking ovflist), but the "unnecessary" test protected us from seeing it in practice.

No. Bug was introduced by the removed test. Thomas contacted me telling that such test was un-needed, and at first sight it looked OK to drop it (since during the loop time the callback inserts into 'ep->ovflist'). But we didn't notice that a premature loop exit might leave items linked to 'txlist', that the splice below takes care of re-insert them. You can speed up things by sucking the patch directly, w/out waiting Andrew's pull.

- Davide